Reconstructing directed gene regulatory network by only gene expression data.

Area of Science:

• Bioinformatics
• Systems Biology
• Computational Biology

Background:

• Gene regulatory network inference is crucial for understanding biological activity.
• Current methods often rely on gene expression data but struggle with determining regulatory direction.
• Additional data, like genetic variations or knockout experiments, are frequently unavailable for human tissues.

Purpose of the Study:

• To develop a novel method, Context Based Dependency Network (CBDN), for inferring gene regulatory networks with directional information.
• To utilize only gene expression data, overcoming limitations of methods requiring additional experimental data.
• To identify important regulators within the inferred networks.

Main Methods:

• CBDN computes regulatory direction by evaluating changes in gene expression dependencies when conditioning on a source gene.
• It extends the data processing inequality to differentiate direct and transitive gene relationships.
• The method defines and detects two types of important regulators based on their influence across the network.

Main Results:

• CBDN successfully infers gene regulatory networks with directionality solely from gene expression data.
• The method outperforms state-of-the-art approaches in accuracy, even when directionality is considered.
• CBDN identified key regulators, including TYROBP linked to Alzheimer's disease and ZNF329/RB1 associated with brain tumor gene signatures.

Conclusions:

• CBDN efficiently infers gene-gene interactions and their regulatory directions using only gene expression data.
• The inferred networks aid in identifying critical regulators for complex diseases.
• This approach provides a valuable tool for biological network analysis without requiring supplementary experimental data.